The Commons Lab Thesis

The Contest for Sovereign Technology

The most consequential question about AI infrastructure right now is purpose, not capability. The defaults available today divide into two surveillance-oriented stacks: a Western enterprise-and-state-security stack optimized for targeting and operational control, and a Chinese walled-garden stack optimized for state-aligned consumer scale. A third pattern joined them in 2026: capture wearing sovereignty language. Vendors now market proprietary stacks as the sovereignty solution while keeping the control layer, the ontology, the permissions, and the audit trail, in their own hands and under their own government’s jurisdiction. Sovereigns who want none of the three, particularly middle-income democracies, are looking for infrastructure that doesn’t export the assumptions of its country of origin. Gartner projects that by 2028, 65% of governments worldwide will impose technological sovereignty requirements specifically to protect against extraterritorial regulatory interference. That is the gap. AI infrastructure can be designed for human freedom, happiness, meaning, and purpose. The demand for that design no longer comes only from technologists. Pope Leo XIV’s first encyclical, released in May 2026 on the 135th anniversary of Rerum Novarum, argues that AI must serve humanity rather than concentrate power, and that technology is never neutral because it “takes on the characteristics of those who devise, finance, regulate and use it.” The coordination layer underneath (credentials, payment rails, governance protocols, attribution systems that connect AI to humans, sovereigns, and each other) is where that third answer gets built, or fails to.

The institutional architecture that governed global coordination for eighty years is coming apart, in a cascade that has accelerated over the past eighteen months. USAID was dismantled, its $68 billion annual disbursement apparatus shuttered, with the Lancet projecting 9.4 million additional deaths by 2030 if current funding trends hold. The UK, Germany, France, and Switzerland followed with their own ODA cuts, redirecting budgets to defense. In late February, US and Israeli strikes on Iran effectively closed the Strait of Hormuz, disrupting 20% of global seaborne energy. The conflict also produced the first kinetic military attacks on hyperscale cloud infrastructure in history: Iranian drones struck AWS data centers in the UAE and Bahrain, taking banking, payments, and enterprise services across the Gulf offline. Sovereign digital infrastructure shifted overnight from an economic preference to a military necessity. The rupture now runs through the alliance itself. In January, European heads of government met in emergency session in Brussels; the theme was how to manage a breakup with America. The follow-through is procurement. Governments from France to the Netherlands are stripping American software out of state systems and moving civil servants onto European open source, while committing hundreds of billions to European AI companies, data centers, and space firms. The studies these governments commissioned tell you what sovereignty means once it becomes a requirements document: where would our data live if relations deteriorate, where would our payments clear, and would our American-made weapons still function without Washington’s authorization. Dependency audits are now standard behavior among America’s closest allies, and a dependency audit is a requirements document for whoever builds the replacement. The post-WWII assumption that a single dominant power would underwrite global stability through institutions and aid has ended.

Two technological shifts are converging on the same structural question: who controls the infrastructure that governs how value flows?

AI agents are becoming autonomous economic actors.According to KPMG’s Q1 2026 AI Pulse Survey, over half of organizations are now actively deploying AI agents, with average AI spending of $207 million per organization over twelve months in the US, doubled year-on-year. Multi-agent systems are negotiating, transacting, and making allocation decisions without human approval. Five competing agent payment protocols launched within 90 days in early 2026: Visa, Google, Coinbase, Stripe/OpenAI, and PayPal each racing to become the default settlement layer for machine-to-machine commerce. The governance infrastructure for these systems does not yet exist.

Sovereign AI has shifted from aspiration to policy. Over 100 countries adopted AI sovereignty declarations at the February 2026 AI for Developing Countries Forum, South Korea announced $735 billion in sovereign AI investment, the African Development Bank and UNDP launched a dedicated AI infrastructure initiative, and mid-sized economies are forming compute alliances and regional GPU clusters. Every one of them needs coordination, governance, and funding infrastructure they can control, audit, and fork.

These forces converge on a single structural need: open-source coordination infrastructure for the intelligent economy. Infrastructure that governs how value flows when centralized institutions are failing, autonomous agents are proliferating, and national sovereignty increasingly depends on digital infrastructure independence.

Commons Lab backs the founders who build that infrastructure. First, though, the word itself: what makes technology sovereign.

What Makes Technology Sovereign

Sovereign technology is defined by a verifiable set of properties.

The sovereign stack is the full infrastructure layer that enables nations, communities, and individuals to maintain agency in the intelligent economy. Own your models. Own your data. Own your compute. Own your robots. Own your exit. Have things work on-device, at the edge, under your control. For a nation, this means digital independence. For a community, it means self-governance. For an individual, it means access to intelligence and the ability to create and exchange value without depending on intermediaries or foreign platforms. Projects across this stack are already emerging: decentralized training networks like Bittensor and Gensyn, sovereign robotics companies like Solo Tech, edge AI, proof of control mechanisms, on-device inference. The movement toward sovereignty at every layer of the technology stack is accelerating.

Exit is the item most sovereignty pitches leave off the list, so it gets its own ladder. The market now distinguishes structural assurance from contractual assurance: guarantees enforced by physics and cryptography versus guarantees enforced by a contract and a vendor’s continued goodwill. Apply that distinction to the control layer, where an institution’s ontology, permissions, and audit history accumulate, and four rungs appear. Open protocols with cryptographic state and capture-resistant governance sit at the top: leave whenever you want, take everything, keep interoperating. Self-hosted open source with open formats sits below: take your code, data, and workflows, give up the shared network. Then proprietary systems with negotiated export rights, where exit is a migration project resting on contract terms. At the bottom, proprietary systems with no portability, where exit means abandoning the institutional knowledge you accumulated inside someone else’s product. One test cuts across every rung: jurisdiction. A vendor answers to its government, so servers on your soil running a foreign company’s proprietary stack are contractual assurance wearing a structural costume. The Swiss Army’s lawyers worked this out on paper. Every Fable 5 customer learned it in production.

With open protocols, exit is real: a sovereign that leaves keeps its code, its state, and its workflows, and gives up the shared network. That is losing the neighborhood. With a proprietary control layer, exit forfeits the institution’s own accumulated ontology. That is losing the house. Staying for network effects is a choice; being unable to leave is capture, and the distance between those two conditions is our focus.

The Money Precedent

The post-1945 assumption extended beyond military alliances and development aid. It included a single, ostensibly neutral financial infrastructure: SWIFT, correspondent banking, dollar clearing. That assumption ended in 2022, when the US and EU froze $300 billion in Russian central bank reserves and disconnected Russian banks from SWIFT. The action was unprecedented. It demonstrated that the financial infrastructure the world treated as neutral plumbing was a weapon available to whoever controlled it.

The response has been a scramble for alternatives, rather than a return to multilateralism. China’s CIPS clears trillions annually, India’s UPI runs over 13 billion transactions a month, and Brazil’s Pix now moves $557 billion monthly, more than Visa and Mastercard combined, proof that state-backed open payment infrastructure can capture total domestic market share. And in April 2026, Iran’s Revolutionary Guard began collecting tolls on oil tankers transiting the Strait of Hormuz, payable in cryptocurrency and Chinese yuan, codified into law by Iran’s parliament as the “Strait of Hormuz Management Plan.” A sanctioned state is now operating blockchain-based revenue collection at the world’s most critical energy chokepoint, the first durable deployment of stablecoin infrastructure as a state revenue mechanism in global trade.

But the escape from dollar hegemony is not what it appears. Shanaka Anslem Perera’s analysis of what he calls the Maduro Paradoxreveals a deeper mechanism. Under comprehensive US sanctions, Venezuela did not pivot to yuan or barter. Eighty percent of its oil sales shifted to centralized dollar-denominated stablecoins backed by US Treasury bills and freezable by a single private company. Within days of Maduro’s capture in January 2026, a stablecoin issuer froze $182 million in wallets reported to be linked to Venezuelan oil transactions, without judicial process. A single private intermediary exercised enforcement capability the entire SWIFT network could not match. The mechanism that was supposed to enable sanctions evasion became the mechanism of sanctions completion. A broader pattern has taken hold. Newer dollar-backed stablecoins have embedded the financial interests of politically connected actors directly in the infrastructure that US policy is legislating into dominance. The line between public monetary policy and private value capture has effectively dissolved.

Iran may be walking into the same architecture.

This is no accident. The GENIUS Act codified private stablecoins as the preferred digital dollar infrastructure. A provision buried in a housing bill banned the Federal Reserve from issuing a Central Bank Digital Currency. And in April 2026, FinCEN and OFAC jointly proposed rules treating stablecoin issuers as financial institutions with authority to freeze and block transactions globally and unilaterally. The European Parliament calls this strategy “cryptomercantilism.” The IMF’s Hélène Rey describes it as “privatization of seigniorage.” Both assessments confirm that what appears to be organic market adoption is industrial policy designed to extend dollar hegemony through private infrastructure.

This creates a false binary that the current geopolitical moment is forcing into the open: either financial infrastructure remains under Western control (SWIFT, and now its stablecoin successor), or it falls under alternative state control (CIPS, IRGC-administered toll systems). Both options concentrate power, both can be weaponized, and neither serves the billions of people and communities caught between competing great powers.

The market has started conceding the point. In June, more than 140 incumbents, Visa, Mastercard, Stripe, Google, BlackRock, and dozens of global banks among them, announced Open USD, a consortium stablecoin with shared reserve income and governance by a board of partners. Their own announcement names the problem this thesis names: businesses have little recourse when a single issuer controls the rails. The answer stops one layer short. The reserves sit at US financial institutions under US regulatory requirements, and US stablecoin law requires issuers to retain the ability to freeze and seize tokens on lawful order. A consortium distributes the yield and the roadmap; it does not distribute the switch. Neutral among a hundred and forty members is not neutral for the eight billion people outside the boardroom. The Maduro Paradox scales: what was true of one issuer is now true of a bloc.

There is a third option.

Developers, researchers, and policymakers across 50+ countries are already building it: infrastructure that is sovereign, interoperable, and capture-resistant at once. The technical primitives exist and the institutional demand is arriving; what’s missing is the coordination layer that connects them.

This is the d/acc response: infrastructure that is open-source, auditable, and resistant to capture by any single state or corporation. Permissionless here has a narrow meaning: no single actor can freeze the system, surveil it without accountability, or weaponize it against the populations who depend on it. This is the infrastructure Commons Lab backs. The design constraint is not neutrality, which has proven illusory, but sovereignty at every layer, from the individual wallet to the national payment rail. To be clear: this is not sanctions evasion infrastructure. Decentralized protocols can be compliant everywhere precisely because they are modular: nodes run jurisdiction-specific compliance parameters, ZK proofs demonstrate compliance by default without exposing user data, and selective disclosure under judicial oversight follows established mutual legal assistance processes, so legitimate law enforcement proceeds while unilateral authoritarian action is structurally prevented. Compliance is verified at the edges before anyone enters a pool and continuously inside it, the design lesson of Tornado Cash, whose unified pool without internal verification became legally toxic. The dual-use risk is real, and the defense is compliance built into the protocol layer (rather than weaker infrastructure). The goal is infrastructure harder for hostile actors to exploit than centralized alternatives.

The AI Front

The AI safety conversation treats redistribution and governance as policy preferences; this thesis treats them as engineering requirements and as preconditions for functioning markets.

Luke Drago’s Intelligence Curse framework draws an explicit parallel to the resource curse in development economics: countries that discover massive natural wealth often end up with worse outcomes because the wealth concentrates rather than distributes. AI is the new resource. The curse is the same, except the attack surface scales with the technology itself. In a petrostate, the displaced population has limited tools. With AI, the same capabilities generating trillions in surplus are simultaneously becoming available to the billions of people cut out of that value. Dario Amodei, CEO of Anthropic, warned in January 2026 that sufficiently powerful AI could enable small groups to synthesize biological agents. When the head of a frontier lab says the attack surface extends to biological weapons, the security case for redistribution stops being abstract.

You cannot build a perimeter when the number of people with reason to breach it grows with every advance in the technology they’ve been excluded from. Redistribution is security architecture.

In April 2026, Anthropic proved this with a product decision. Its new frontier model, Claude Mythos, autonomously discovered thousands of zero-day vulnerabilities across every major operating system and web browser, including a flaw in OpenBSD that had gone undetected for 27 years. The capabilities weren’t trained for. They emerged as a byproduct of general improvements in reasoning and code generation. Anthropic chose not to release the model publicly. Instead, it launched Project Glasswing: early access restricted to a coalition of the world’s largest technology and financial companies, plus organizations maintaining critical open-source infrastructure, backed by $100 million in usage credits (Apple, Microsoft, and JPMorgan Chase among them) to patch critical infrastructure before the same capabilities proliferate to hostile actors.

This is the Intelligence Curse operating in real time. The surplus exists. Releasing it broadly would compromise global infrastructure. So the most powerful tool yet built, offensive and defensive simultaneously because finding a vulnerability and exploiting it are the same capability, concentrates in the organizations that are already the most capitalized and best defended, while everyone else inherits the expanded attack surface without the tools to address it. Glasswing is also a warning about the limits of benevolent centralization: one private company now holds working exploits for most of the world’s critical software. That is itself a concentration risk. In June, the warning stopped being hypothetical. On June 9, Anthropic released Fable 5, the first generally available model of its Mythos class. Three days later the US Commerce Department issued an export-control directive citing national security, barring access by any foreign national anywhere in the world, including Anthropic’s own foreign-national employees. Nationality cannot be verified at the API layer, so compliance meant switching the model off for every user on earth, within hours, while the company publicly disputed the basis for the order. The directive was lifted on June 30 and service resumed July 1. The retention terms tell the vendor’s side of the same story: Anthropic’s own launch policy for the model class mandates 30-day retention of prompts and outputs and does not honor existing zero-data-retention agreements, a unilateral carve-out that overrode prior enterprise commitments. The state can throw the switch, and the vendor can rewrite the terms. That is what contractual assurance looks like from the inside. The same month, OpenAI launched its next flagship only as a government-cleared preview for vetted organizations. The precedent now covers both of the market’s leading labs: generally available is a conditional state. Frontier capability concentrated in one company becomes a single switch, and the hand on the switch belongs to that company’s government, regardless of the company’s own judgment. We wrote the April version of this thesis warning about this scenario. Two months later the state completed the argument. The structural response is decentralized, sovereign security infrastructure that distributes defensive capability without centralizing offensive knowledge, because a better monopolist is still a monopolist.

The demand signal stopped being a forecast this year. France’s domestic intelligence service is replacing Palantir with French-built software, Germany is reportedly making the same move, and Mistral is being deployed across the French civil service. In Switzerland, an internal army evaluation concluded that Palantir’s systems could be incompatible with Swiss sovereignty: even with servers on Swiss territory, American law could compel the company to grant access to the data. The evaluators recommended the army forgo Palantir’s solutions. Sovereignty has moved from declaration to procurement requirement, and the incumbents have noticed; the largest of them now packages open-weight models into air-gapped government deployments under a sovereign banner and offers governments free sovereignty reviews. The window this thesis described in April is now being contested in the open, which is what windows look like when they are real.

This is also the business case. Capital allocates efficiently when participants trust the system won’t be captured or weaponized against them, and the Intelligence Curse describes exactly the instability that destroys that trust: regulatory backlash, political upheaval, an exponentially growing attack surface, authoritarian capture. Every one of those outcomes is bad for returns. The infrastructure that distributes power and wealth is a precondition for market function, not a competitor to it.

The loop runs: safety enables deployment, deployment generates surplus, surplus requires redistribution, redistribution reduces the attack surface, and a reduced attack surface keeps markets functional enough to deploy further. Break any link and the architecture fails. Coordination infrastructure sits at the center of the loop, not as a charitable cause but as the stability layer that lets the broader intelligent economy function.

The default deployment model for frontier AI reinforces this concentration. Frontier labs build the most capable models, then wrap them in enterprise tooling designed for organizations with existing data, capital, and integration capacity. A Fortune 500 company deploying AI agents across thousands of employees captures compounding productivity gains. A five-person organization in Nairobi gets the same general-purpose chatbot as everyone else. This is the business model, not a market failure waiting to be corrected. Commons Lab backs infrastructure that breaks this default and distributes capability rather than concentrating it.

But rails that run through state-controlled channels can be weaponized, and AI surplus carries the petrostate risk with a far more capable tool for maintaining control. Open-source, permissionless infrastructure is an engineering requirement for coordination systems that can’t be turned off by the parties who benefit most from concentration.

Everything Commons Lab backs is underwritten by this structural logic: The coordination infrastructure of the intelligent economy must be open-source, resilient, and sovereign. The alternative is concentration that produces an exponentially growing attack surface, or state capture that produces intelligence-powered authoritarianism. Both destroy the market conditions that make technology investment viable.

The Agent Economy

The next major infrastructure cycle is the economic plumbing of autonomous systems, and it looks nothing like another consumer application layer.

Consider what agents need to transact at machine speed: access to payment rails (no bank account applications, no per-transaction authorization), programmable escrow (contracts that release funds on verified task completion), composable identity (wallets and reputation that accumulate across services), and auditable decision trails (verifiable records of what was allocated, why, and what resulted). Not every interaction settles on-chain: high-frequency micro-transactions batch through off-chain channels and settle periodically to a base layer that provides finality and dispute resolution. Traditional financial infrastructure was designed for humans operating at human speed with human trust signals. Agent-to-agent commerce needs infrastructure that is programmatic by default, AI-native by design, and verifiable without intermediaries.

Crypto-native rails are the most logical foundation for this. On-chain smart contracts, stablecoin settlement, and programmable governance are already designed for exactly the properties autonomous agents require. The regulatory ground has settled underneath them: federal stablecoin law in the US, MiCA across the EU, and licensing regimes advancing through most of the G20.

But the coordination and governance layer these systems need is still nascent: agent-to-agent payments, machine-readable compliance, and frameworks governing what agents can do are not yet deployed standards.

Agent marketplaces are themselves coordination infrastructure. When autonomous agents need to discover each other, negotiate, transact, verify outcomes, and resolve disputes at machine speed, the marketplace layer is governance architecture.

Agent-mediated governance may be an even larger opportunity. In the Simocracy experiments at Frontier Tower, community leaders created AI agents representing their priorities, and the agents deliberated among themselves to allocate shared resources. Humans define values and preferences; agents handle the coordination overhead. The applications extend to any group whose collective decisions are bottlenecked by time, attention, or apathy. Voter turnout in US local elections averages around 20%. The problem is coordination cost, and agents collapse it. The infrastructure to make this work safely (transparent deliberation, verifiable preference representation, accountable allocation) is coordination infrastructure.

Commons Lab’s thesis is that the defining structural opening of the agent economy is already visible: it is bifurcating into two enterprise stacks, and the split is durable. In the West, x402, AgentCore Payments, AP2, and pay.sh are standing up agent-to-API commerce on stablecoin rails: open protocols in name, but value capture concentrated in Coinbase, AWS, Google, Stripe, and the Solana Foundation. In China, Alipay’s Agentic Commerce Trust Protocol, integrated with Qwen and Taobao Instant Commerce, processed over 120 million AI-agent transactions in a single week in February 2026, far beyond the volume reported by any Western agent payment protocol to date, but locked into Alipay’s ecosystem and difficult to access for non-Chinese builders. Chinese open-source AI labs (Qwen, DeepSeek, and others) are simultaneously excluded from the Western crypto-rail upside. The result is two operationally incompatible agent economies, neither building the universal coordination primitives the technology requires: agent identity, contribution attribution, public-goods funding for shared infrastructure, and credibly neutral governance. The entity that funds and stewards the neutral, transnational, public-goods layer will shape how agent-mediated coordination is governed over the next decade.

Commons Lab backs founders building the neutral layer underneath both stacks. The priorities are concrete and already mapped: agent identity that works across jurisdictions, middleware between competing payment protocols so no single stack becomes a chokepoint, and capture-resistant settlement rails that interoperate with sovereign monetary systems. The full opportunity map is at commonslab.ai/stack. The work is keeping the layer between the walled gardens public, neutral, and credibly governed.

These founders build the governance, compliance, and allocation infrastructure that the agent economy will run on, and Commons Lab creates the conditions for adoption by embedding them inside institutional deployment environments.

Where Demand Starts: The Beachhead

Development finance is the beachhead because the demand signal is loudest and the institutional buyers are already engaged. The institutional order is collapsing fastest there, the need for transparent coordination is most acute, and the same infrastructure serves every context where value needs to flow transparently.

The global development finance system moves $200 billion annually through multilateral and bilateral channels. While multilaterals report 5-15% overhead, cumulative losses across multi-tier subcontracting, FX spreads, and corruption mean less than half of each dollar reaches beneficiaries. In the worst cases, less than 10% reaches local organizations directly; the rest is absorbed by intermediaries and revolving-door procurement networks capturing margin at every tier. US ODA for 2026 is projected to fall 56% from 2023 levels. The UK cut from 0.5% to 0.3% of GNI. The assumption that other donors would step up to compensate has proven wrong. The development funding landscape is reorganizing around whoever can deploy transparently within the next 18 months.

The buyer on the other side of this collapse is government, and governments are becoming major AI buyers in their own right. Gartner forecasts global AI software spending in the government market to grow from $41.6 billion in 2024 to $70.6 billion by 2027, and projects that by 2028 at least 80% of governments will deploy AI agents to automate routine decision-making. The question is whether they buy off-the-shelf enterprise stacks optimized for surveillance and control, or commission infrastructure designed around citizen agency. Govtech AI is the market wedge, and development finance is govtech through the side door: the same sovereign buyers, with more urgency and fewer incumbents.

A second demand vector is emerging from digital-state pioneers integrating AI into citizen services. Estonia’s Bürokratt has woven AI-powered virtual assistants across 18 government agencies, building on the X-Road data exchange platform, and is now extending toward cross-border interoperability with other national assistants. Singapore’s GovTech has deployed Pair, a secure LLM environment for public officers, and AI Verify, a governance testing framework, and published one of the first government-issued Agentic AI Primers in April 2025. These deployments raise coordination questions the global AI policy conversation has not yet answered: who controls the model, who audits the outputs, who gets remediated when the system errs, how agent capabilities transfer across borders. The countries piloting these systems are doing the implementation work in advance of the protocol layer that should support it. That gap is where Commons Lab meets a sovereign buyer with budget, urgency, and willingness to pilot.

Programmable infrastructure changes the cost structure of moving value from funder to beneficiary. Transparent ledgers replace bolted-on reporting. Smart contracts encode allocation rules and milestone verification. Stablecoin rails cut the cost of moving a dollar from 3-7% to under 1%. The World Food Programme already reaches 4 million beneficiaries monthly using permissioned blockchain ledgers, proving the operational efficiency of the architecture; the next phase is migrating these siloed systems to open, interoperable rails. AI agents can deliver treasury management, impact documentation, and fraud detection at near-zero marginal cost, replacing the consultant infrastructure that intermediaries previously required. Physical-world verification still costs money, but community attestors, satellite imagery, and sensors cost orders of magnitude less than the auditor apparatus they replace, and they produce cryptographically verifiable records rather than PDF reports. Outcomes-based mechanisms change what gets rewarded. The historical bottleneck in outcomes-based finance has been twofold: pricing outcomes accurately and finding buyers. Crypto-native infrastructure addresses both. Tokenized impact certificates open participation to global capital that traditional gated instruments exclude, and mechanism design from the public goods funding space (retroactive rewards, prediction markets, prize competitions) aligns buyer, funder, and implementer without bilateral negotiation. The result: disintermediation compresses cumulative fees from 30-60% to under 10%. The UBS SDG Outcomes Fund validated the model, raising $100 million in blended capital and delivering $13.5 million in verified outcome payments across health, education, and employment programs. Commons Lab is developing a Hypercert Outcome Bond specification that combines blended capital stacks with first-loss tranches, multi-metric outcome verification, and marketplace settlement into a replicable standard for outcomes-based financing on-chain.

Cost recovery is the smaller win; the larger part is a structural shift from extractive to positive-sum dynamics. Today intermediaries are funded to manage problems and lose their budgets when problems are solved. Transparent, programmable allocation inverts that: communities managing funds through open treasuries maximize impact because future allocation depends on demonstrated results, and the cycle compounds, more value reaching beneficiaries, better outcomes, more capital.

Across the programmable portion of the $200 billion in annual development flows (bilateral grants, multilateral disbursements, humanitarian cash transfers), this represents tens of billions in recoverable and newly created value annually.

Fixing the plumbing of development finance is the entry point to something much larger.

The larger opportunity is economic participation.

As work automates and access to intelligence becomes universal, 4-5 billion people in the Global South gain access to capabilities that previously required expensive expertise. On-device models running at the edge give a farmer in Mombasa the same quality of guidance a consultant in Washington bills $500/day for. The missing piece is the coordination infrastructure that connects these participants to global value networks; intelligence itself is becoming abundant. The Global South is the largest untapped source of human creativity and economic participation on the planet, not a population waiting to receive aid, and the intelligent economy gives us the tools to unlock it.

The same primitives apply to municipal budgets, climate finance, humanitarian response, and agent-to-agent economic coordination. The infrastructure our founders build in Africa gets refined for autonomous systems everywhere.

The Adoption Model

MOSIP is the adoption model. SWIFT proved shared financial infrastructure can operate across 200 countries and sustain itself through transaction fees, but SWIFT is proprietary and geopolitically captured. Russia’s exclusion showed that “neutral” infrastructure controlled by one bloc isn’t neutral.

MOSIP (Modular Open Source Identity Platform) and X-Road proved the alternative: open-source infrastructure that governments adopt because they can control, audit, and fork it. MOSIP has reached 29 countries with over 100 million digital IDs. Africa has become the world’s largest digital identity test case, with 16 major schemes covering 300 million people. Ethiopia’s MOSIP-based Fayda ID is enrolling one million people per week and has been included in the country’s sovereign wealth fund portfolio, a first globally. What India did for sovereign identity, the next decade requires for sovereign AI.

But building sovereign infrastructure does not guarantee it gets used. India committed over $200 billion to sovereign AI and provisioned tens of thousands of GPUs. Utilization runs at 22%. Enterprises choose hyperscalers because the ecosystem is stickier and venture capital bundles cloud credits. MOSIP’s own president acknowledged “very meagre” commercial returns after eight years of deployment. The lesson is that sovereignty without coordination infrastructure is expensive symbolism. The coordination layer makes sovereign systems usable, interoperable, and commercially viable; without it, countries build hardware that sits idle. Commons Lab is focused on the activation layer that makes sovereign infrastructure work, and activation layers come in two kinds: the kind you rent and the kind you can fork. Only the second passes the exit test this thesis opened with. We back founders building the second. Sovereign compute remains under sovereign control. The coordination layer helps it interoperate with other sovereign systems without surrendering governance to a borderless network. The binding physical constraint for Global South sovereign compute is energy, not hardware alone. The coordination layer routes inference and batch jobs to follow grid surplus: geothermal in Kenya, solar in the Gulf, hydro in the Nordics, turning a constraint into an arbitrage opportunity. Not every sovereign will accept this. States driving the splinternet will reject open coordination to maintain domestic control. The target market is mid-sized democracies and allied blocs that must coordinate to compete with superpowers but cannot afford to surrender sovereignty to a foreign tech stack.

MOSIP’s meagre commercial returns reflect the absence of protocol-level value capture rather than any structural limit of open-source infrastructure: Ethereum is also foundation-stewarded, and it captures value through gas fees and algorithmic supply reduction. Red Hat achieved a $34 billion acquisition by providing enterprise support for open-source Linux. Open-source digital public infrastructure can be monetized ethically when mechanism design aligns network participation with value accrual. This can happen through transaction fees, enterprise licensing, or protocol revenue in whatever currency the sovereign mandates.

We back founders building MOSIP-style coordination infrastructure with sustainability built in: open-source protocols with revenue capture through enterprise support, transaction fees, or value-added services. Every country building a sovereign AI stack also needs the coordination layer that governs how those systems interact with people, institutions, and each other. The addressable market now includes every mid-sized economy building digital independence.

What We Back: The Coordination Layer

Commons Lab accelerates the coordination layer of the sovereign stack. Sovereign technologies are only useful if they can coordinate with each other: transact, allocate resources, verify compliance, resolve disputes, and govern shared infrastructure. We back founders building that coordination infrastructure. The categories run in dependency order, from the settlement floor to the agent frontier: each layer relies on the ones beneath it.

Capture-Resistant Settlement. Sovereign money rails: settlement that interoperates with sovereign monetary systems. Every solution now on offer, single issuer or consortium, keeps the freeze switch, and a sovereign cannot run funding, compliance, or agent commerce on rails that can be shut off from outside its own jurisdiction. Settlement can happen in CBDCs, authorized stablecoins, or fiat; what the layer must guarantee is that no single actor can freeze it and that exit passes the fork test. Revenue through transaction fees and enterprise integration.

Privacy-Preserving Compliance and Identity. The enabling layer for institutional adoption and the trust layer for the intelligent economy. ZK proofs, homomorphic encryption, and selective disclosure let you prove compliance without exposing underlying data.

Privacy also produces structural defensibility. In commoditized blockspace, protocols holding private state generate lock-in that transparent chains cannot replicate: users who enter a private zone cannot cross back out without exposing the metadata they entered to protect. The protocols that establish this first can capture the bulk of institutional and sovereign deployments.

Equally critical: proving you are human or proving you are an agent. You cannot run agent-mediated governance, programmable allocation, or compliant transactions without knowing what kind of entity you’re interacting with, so proof of personhood, proof of agency, and privacy-preserving credentials that satisfy regulators without surveillance are core coordination primitives. This does not mean anonymity: the compliance architecture from the Money Precedent applies, with selective disclosure under judicial oversight satisfying legitimate law enforcement. Protocols that cannot satisfy either standard will be banned. This is the clearest revenue story: compliance and identity verification are cost centers institutions must pay for, and the need intensifies as agents transact at machine speed.

Programmable Funding Mechanisms. Infrastructure for allocation, disbursement, and impact verification: outcomes-based financing, agents that represent the will of individuals and communities. These mechanisms run on the settlement and identity layers before them. Without capture-resistant rails and verified counterparties, they inherit the freeze switch. As AI agents take on treasury functions and allocation decisions, this becomes the governance layer they operate within. Revenue through transaction fees, SaaS licensing to institutions, or protocol revenue share.

Distribution Infrastructure and Data Sovereignty. Last-mile infrastructure that governments and multilaterals use to move funds: open-source, sovereignty-preserving, interoperable with the verification and compliance layers. Data sovereignty is negotiated through cryptographic primitives: individuals and communities control their data, decide what enters the shared commons, and negotiate rights and rewards. Frontier labs are shifting toward synthetic data for pre-training, but sovereign AI stacks depend on high-integrity local data that changes continuously: crop conditions, health outcomes, economic activity, community needs. This is a continuous ground-truth stream that retains recurring value because it cannot be synthesized. The distribution layer becomes the data collection layer, and the communities generating that data become participants in the value it creates.

Agent-Native Coordination. AI agents interacting with economic systems need governance architectures: agent-to-agent payment systems, observability frameworks, proof of control mechanisms, and protocols that determine what agents can do with whose resources. Enterprise control layers can already govern agents inside a single institution’s perimeter, and that is where they will stay, because governing agents across perimeters requires something no vendor can sell: a referee both counterparties trust. When an agent spawned under one jurisdiction transacts with an agent spawned under another, neither side accepts the other’s vendor as judge. And the cross-boundary economy is already here: roughly 176 million agent transactions were tracked between May 2025 and April 2026, averaging 31 to 48 cents each, which puts about three quarters of them below the fee floor where card rails are viable. Machine-scale commerce has arrived under human-scale infrastructure’s minimum transaction processing fee. The legal surface area is unresolved: if an autonomous agent executes a trade or misallocates funds, liability must attach to a legal entity. Blockchain-based provenance provides the technical answer: a record of who spawned an agent, under what constraints, and with what authorization creates an auditable chain of responsibility, even when agents spawn other agents autonomously. Provenance alone is not sufficient: agent marketplaces and institutional APIs must require collateral and/or a verified legal entity before granting access. This includes agent marketplaces, where trust is programmable and accountability verifiable, and the agent-mediated governance described earlier.

As AI models generate interfaces on the fly, the value capture point shifts. Products competing on user experience face commoditization from model providers who replicate interfaces at marginal cost. Durable value migrates downward to the infrastructure that cannot be replicated at marginal cost: identity and credential layers, data sovereignty systems, agent coordination protocols, and privacy-preserving compliance rails. Commons Lab backs founders building below the application layer, where defensibility compounds.

Previous generations of open protocols (TCP/IP, SMTP, HTTP) failed to capture value at the protocol layer; value aggregated to application-layer monopolists instead (Google, Meta, AWS). This is the core of Aggregation Theory, and it is the default objection to investing in open-source infrastructure. Crypto-native protocols break the pattern by embedding economic mechanisms directly in the protocol that let the infrastructure layer capture value proportional to usage. Ethereum, stewarded by a nonprofit foundation, generated $2.48 billion in fees in 2024, more than any other blockchain. That is the structural innovation that makes open-source coordination infrastructure investable at venture scale. The objection: institutions use SWIFT because they want a centralized intermediary to bear legal liability when things break, and on-chain governance today is higher-friction than a bank account. The protocols we back solve technical problems; portfolio companies must also solve the sociological and legal ones. The enterprise wrapper answers both: it provides SLAs, compliance gating, and liability strictly for the deployment instance the institutional client uses. The wrapper is capturable by design: if a government orders censorship, it complies, and clients lose that access point. The protocol survives through client diversity. Its value is that it cannot be killed, even when individual access points are captured.

How We Execute

Our UNDP AltFinLab partnership has evolved from a conference panel into exploring opportunities to deploy funding mechanisms and blockchain infrastructure with UNDP across Africa. The work is designed to bring international and regional founders from across Africa to live deployment challenges drawn from UNDP Country Office problem statements: digital identity, cash transfer optimization, climate finance verification, and programmable allocation for humanitarian response. Over five years, Funding the Commons has built a 24,000+ person developer and researcher network across four continents and produced over 200 prototypes through hackathons and three dedicated residencies. That community is the talent and research base these institutional programs draw on. This partnership is also designed to function as a procurement accelerator. Sovereign and multilateral buyers operate 18-36 month procurement cycles that seed-stage ventures cannot survive. The residency model deploys founders directly inside institutional environments, proving the technology works before formal procurement begins. By the time an RFP issues, the technology is already operational. When an institution issues an RFP requiring compliance with open standards, the portfolio companies that helped build them carry a strong implementation advantage. Any vendor can bid. The Tor Project approached us to help diversify their funding base after losing US government grant funding, connecting privacy protocols to their 3 million daily users.

Our residency model brings Global South founders to major tech hubs, giving them direct access to the capital markets, networks, and technical talent concentrated in places like San Francisco. We’ve built relationships with frontier tech communities in these cities, including Frontier Tower (SF, planning to expand to London and NYC), where international founders embed alongside frontier tech builders of all stripes. The proximity produces results no remote program can match. But the value flows outward: founders access capital and mentorship in SF, then build and deploy at home. Our March 2026 hackathon at Frontier Tower validated builder appetite for this approach, with multiple teams building agent governance prototypes against real community coordination problems.

The pipeline is already active. FtC’s residencies have partnership conversations with major protocol ecosystems for chain-exclusive builder deployments, alongside UNDP institutional co-production. Protocol sponsors fund the residencies; builders deploy real applications on specific chains; institutional partners provide the deployment environment and credibility. The builders deploying stablecoin rails and programmable funding mechanisms today are creating the same coordination primitives that will underpin agent-native systems and sovereign infrastructure tomorrow. Funding the Commons runs conferences, hackathons, and residencies to serve builders and the public-goods field on their own terms. Commons Lab is an adjacent incubator that founders from FtC’s community can work with, and we back the strongest of them.

Each venture enters incubation matched with an institutional pilot partner, drawing on relationships built through Funding the Commons and the residency model that embeds founders directly inside deployment environments. Resources scale with demonstrated traction, not pitch decks.

The pipeline compounds: Funding the Commons (conferences, hackathons, residencies) → Commons Lab (incubation, stipends, pilot partners) → the planned Commons Fund (seed investment in portfolio companies). Each entity has its own mission, and a founder moves through the stages by choice. By the time one reaches readiness, real deployment has already tested the work.

When we say we back founders, this is what we commit: stipends during incubation, structured mentorship from our advisory network, direct introductions to institutional pilot partners, and a path to seed investment through the planned Commons Fund. We take asset positions with performance-based vesting; founders own their companies, and we earn our position by helping them succeed.

Alumni from FtC programs have become Ethereum Foundation Next Billion Fellows, won the Rainforest X-Prize, and raised follow-on venture capital from major protocols.

Gane represents pre-formation advisory work now held as portfolio equity. Gane is a last-mile distribution network connecting 80,000 users across Latin America to crypto wallets, on-device AI, and basic financial services through subsidized mobile connectivity. Users engage with educational content in exchange for free data and minutes, plus onboarding to stablecoin remittances and programmable funding flows. Gane is the distribution infrastructure category in practice.

Commons Lab is pre-first-cohort. The UNDP partnership we are exploring across Africa is the intended inflection point: the first cohort would be recruited against real deployment problems. The venture categories in this thesis are what tends to emerge from that work. Target outcomes: 1-2 ventures and 4-5 real-world pilots. This is the proof-of-concept for the full pipeline operating with institutional partners.

The pipeline’s defensibility is structural. Incubated founders contribute stakes to a shared portfolio managed by Commons Lab. In return, they receive exposure to the full portfolio index through a single vehicle, not bilateral cross-holdings. Every participant’s success compounds everyone else’s. Portfolio exposure vests based on ongoing ecosystem contribution, preventing free-riding by founders who exit early or underperform. The portfolio can hold equity, tokens, outcome bonds, or other instruments depending on each venture’s structure. Mentors, institutional partners, and ecosystem contributors earn fractional positions in the same index by providing value to portfolio founders. The incentives are self-reinforcing: every stakeholder benefits when any founder succeeds. As builders accumulate portfolio positions, reputation, and relationships within the network, switching costs compound.

The community layer reinforces the economic one. Commons Lab founders embed in physical residency environments across a global network of partner locations, and the relationships, local knowledge, and deployment experience they accumulate there are often non-transferable. Combined with portfolio exposure and institutional pilot access, the network produces switching costs that compound across economic, relational, and reputational dimensions. The longer a founder participates, the more expensive it becomes to leave. The community becomes the moat.

White Space

The coordination layer of the intelligent economy does not have a dedicated incubator. Adjacent funds validate pieces of the opportunity without occupying the intersection. GovTech Fund ($50M, portfolio raised $500M+ in follow-on from General Catalyst, a16z, KKR) and Urban Innovation Fund ($200M+ AUM, produced a unicorn in Jeeves) prove that government and institutional buyers pay venture-scale premiums for infrastructure. Neither sits where coordination infrastructure, the agent economy, institutional deployment, and the sovereign stack overlap.

The incumbent co-option risk is real, and in 2026 it stopped being hypothetical. Within two weeks in July, three of the largest platform companies on earth published the core of this thesis. A consortium of more than 140 financial incumbents, Visa, Mastercard, Stripe, Google, and BlackRock among them, launched Open USD, a stablecoin marketed as open infrastructure under neutral governance, with agentic commerce among its launch uses. Palantir published a fifteen-step white paper, Institutional Sovereignty in the Age of AI, arguing that models are commodities, durable value lives in the layer around them, and structural assurance beats contractual. Days later, Microsoft’s Satya Nadella namedwhat he called the Reverse Information Paradox: that a company using AI pays twice, once in money and once in the proprietary knowledge it reveals just by using the model, with the asymmetry compounding toward whoever owns the learning loop. When the vendors selling the stack, the coin, and the cloud all agree the buyer is the one exposed, the diagnosis is no longer contrarian. However, their prescriptions are revealing. Each answer is a boundary the buyer rents from the vendor proposing it: Palantir’s control layer, the consortium’s governance, Nadella’s tenant. A boundary you rent cannot protect you from your landlord. Open, neutral, sovereign: the vocabulary of this thesis is now the marketing language of the largest vendors and consortiums on earth, and what the language omits is revealing. The Palantir framework runs out one layer short of the product it sells; its assurance ladder covers compute and stops there, and a proprietary control layer sits on the contractual rungs of its own scale. The consortium coin shares the float and keeps the freeze button.

The sovereignty paradox is the inverse objection: if the protocols are truly forkable, won’t every country just run its own instance and destroy the network effects required for venture returns? You can fork the code. You cannot fork the network or the composability. A country can run a forked Treasury OS on an air-gapped intranet, but it loses two things simultaneously: the network effects (user communities, liquidity pools, developer ecosystems that took years to build) and the composability (interoperability with the compliance layer, the outcomes marketplace, the agent governance protocol, and every other sovereign running the shared standard). The value of coordination infrastructure is in the stack and the network, not any single component. Forking one layer means rebuilding all of them and convincing everyone else to switch. Staying is a choice, being unable to leave is capture. Governance of the base protocol must be resistant to capital-weighted capture. If a sovereign wealth fund can buy 51% of governance weight on the open market, decentralization is nominal. Protocol governance must incorporate identity-weighted or quadratic mechanisms that prevent plutocratic takeover. Commons Lab monetizes the activation layer, not the base protocol. Hyperscalers can fork open-source code, but they cannot fork the cryptographic state: the verified history, compliance attestations, and the validator set that provides capture-resistance. The activation layer’s value is access to the live network, not the codebase.

The demand is converging from multiple directions at once. Government buyers need transparent, sovereign coordination infrastructure. The agent economy needs governance, compliance, and allocation rails that don’t exist yet. The sovereign AI buildout is creating new buyer categories across 100+ countries, with sovereign AI spending projected to exceed $100 billion globally in 2026. The collapse of development finance is forcing alternative delivery mechanisms, and 4-5 billion people gaining access to intelligence need coordination infrastructure to connect to global value networks. Just as cybersecurity grew proportionally with the internet economy into a $200B+ industry, coordination infrastructure grows proportionally with the intelligent economy. If AI becomes the largest industry in history, coordination infrastructure for the intelligent economy becomes one of the largest infrastructure categories. The organizations that can deploy transparent, accountable coordination infrastructure within the next 18 months will capture structural position that lasts decades.

We know what types of ventures we back, we have the institutional relationships to deploy them where they matter, and the world has gotten significantly more urgent about the problems we’ve been working on for three years.

David Casey
Founder, Commons Lab
July 2026